Let A:RnRnA:\mathbb{R}^n\to\mathbb{R}^n be an invertible , and equip the space of linear maps with an .

Neumann series lemma. If B:RnRnB:\mathbb R^n\to\mathbb R^n satisfies

A1(BA)<1,\|A^{-1}(B-A)\|<1,

then BB is invertible and

B1=k=0(A1(BA))kA1.B^{-1}=\sum_{k=0}^\infty \bigl(-A^{-1}(B-A)\bigr)^k\,A^{-1}.

Moreover,

B1A11A1(BA).\|B^{-1}\|\le \frac{\|A^{-1}\|}{1-\|A^{-1}(B-A)\|}.

In particular, if BA1/(2A1)\|B-A\|\le 1/(2\|A^{-1}\|), then BB is invertible and B12A1\|B^{-1}\|\le 2\|A^{-1}\|.

Remarks

This lemma is a key linear-algebraic ingredient in the : once Df(a)Df(a) is invertible, Df(x)Df(x) remains invertible for all xx sufficiently close to aa (because DfDf is ).